SKILL.md
SDLC: Intake → Triage → Explore → Plan → Code → Test → Review → Commit → Deploy → Improve
MCP Tools by Phase
| Phase |
Tools |
| Intake |
mcplinearget_issue, mcpnotionnotion-search |
| Explore |
mcpmemorysearchnodes, mcpgithubsearchcode, mcpcontext7query-docs |
| Plan |
mcplinearcreateissue (subtasks), mcpnotionnotion-create-pages, mcpgithubcreatebranch |
| Code |
mcpcontext7query-docs, Read, Write, Edit |
| Test |
Bash (make test, npm test), /quick-test |
| Deploy |
mcpgithubcreatepullrequest, /smoke-test |
| Improve |
mcpmemorycreateentities, mcplinearcreateissue |
Triage Decision
| Change Type |
Path |
| New feature, breaking change, architecture |
OpenSpec → /openspec:proposal |
| Bug fix, config, tests, typo |
Quick → TodoWrite |
Phase Actions
Intake: Fetch Linear issue, search Notion for context Explore: Search Memory for learnings, explore codebase, check related PRs Plan: OpenSpec proposal OR TodoWrite, create branch Code: Follow tasks, use platform rules, mark todos complete Test: /quick-test, add tests for new logic Review: /code-review, /security-scan if auth/data Commit: /commit with type: description Deploy: Create PR, run smoke tests after merge Improve: Store learnings in Memory, create improvement tasks if needed
Quick Reference
# Backend
cd backend && make test
cd backend && make build
# Web
cd web && npm test
cd web && npm run build
# iOS
cd mobile/ios && make test
# Android
cd mobile/android && ./gradlew test
Anti-Patterns
- Coding without exploring → miss patterns
- Skipping Linear context → untracked work
- No Memory capture → repeat learnings
- Big commits → hard to review
